Total mortality in the U.S. elderly population has changed significantly due to declines in heart disease, and stroke and, recently, (> 1990) cancer mortality. Research suggests that, even at advanced ages, morbidity and disability can be modified significantly before death - and influence the diseases responsible for death --through lifestyle changes, cohort experiences, changes in medical technology, and changes in health care access. We will examine temporal and cohort changes in the causes of mortality using several types of data - each with characteristic strengths. One type, multiple cause of death data, will soon be available for a long enough period of time (1968 to 2000) that cohort differences in the multiple conditions reported at death can be examined. Data, from longitudinal population studies, provide information on the temporal relation of morbidity and disability changes for lengthy periods prior to death. This adds a crucial intra-individual temporal dimension to analyses of human failure processes at late ages. To analyze these data we will use statistical estimation strategies designed for analyzing combinations of longitudinal demographic and health survey data sets - each with different statistical and measurement properties. With these data and methods we will examine hypotheses about recent (1990-2000) U.S. declines in overall cancer mortality, i.e., what ages does it affect most, which types of cancer changed most, were changes due to prevention or improvements in treatment, how are estimates of cancer mortality declines affected by trends in circulatory disease death? We will also examine hypotheses about changes in the age dominance of specific conditions, e.g., what types of circulatory disease, and co-conditions, emerge at, say, ages 65 to 75 vs. ages 85+. This will produce insights into changes in human mortality processes at late ages where growing numbers of death will occur in the future as the U.S. population ages and the numbers of nonagenarians and centenarians increase. This is important in forecasting life expectancy and the growth of he extreme elderly U.S. population.